Alcohol consumption is known to be a major risk factor for a wide range of health conditions and injuries. This project focuses on populations in the US and UK, where alcohol consumption is estimated to be responsible for between 3 and 4 deaths in every 100. The patterns of drinking that increase the risk of disease and injury are complex, but include prior history of drinking, average quantities consumed, and how often five or more drinks are consumed in a single occasion. Despite being a harmful substance, alcohol has a long history of use in both US and UK societies and so legislators in these countries have been reluctant to tighten controls on the sale of alcoholic beverages to consumers. In particular, before making any decisions, policymakers are keen to use computer models to help understand what the likely impact of any future policy is likely to be, in terms of how people change their drinking and the subsequent effects on alcohol-related harms. The computer models have helped to bring together some of the evidence on the causes of particular patterns of drinking, and how an intervention (for example, increasing taxes) might change these patterns. They have also been successful in relating these changes in consumption to changes in ill health, hospital admissions, and premature deaths. However the models also have some weaknesses: they do not cover all of the reasons that cause people to drink; they do not fully account for all of the uncertainty in the evidence base in their predictions of policy effect; and they are built in such a way that the predictions they make cannot easily be tested. In this project, our goal is to transform the methods used for alcohol policy models to overcome all three weaknesses. To do this, we need to consider drinking as being a conscious decision or habit of an individual person, but contained within a wider system of influences from other people, institutions, and regulations. Our first specific aim is to test four theories that have ben contended to explain drinking behaviors within five populations (the US as a whole, three states - California, New York and Texas - and also England) over the period 1979 to 2015. The theories are based on personality changes, influence of family, friends and parents, lasting effects of the culture of the era when people grew up, and consumer choice. Our second specific aim is to develop a new systems-based model, integrating useful parts of the theories, and test its ability to represent drinking patterns in the five populations over 1979 to 2015. We will use modern statistical methods to ensure that the models' predictions properly account for uncertainty. Our third specific aim is to extend the systems-based models so they also produce estimates of alcohol-related harms, and we will test these capabilities using alcohol poisoning and liver cirrhosis as outcomes. Our fourth specific aim is to extend the model so that it can estimate the impact of policies, and we will use it to evaluate historical changes in taxation and minimum legal drinking age laws and to make predictions over 2016-2025 for minimum pricing and screening at family doctors for alcohol misuse. Finally, we will make the computer models developed in the project publicly available for everyone to use.